diff --git a/changelog.d/fixes/10465-gemini-cached-tokens.md b/changelog.d/fixes/10465-gemini-cached-tokens.md
new file mode 100644
index 0000000000..0acd31720a
--- /dev/null
+++ b/changelog.d/fixes/10465-gemini-cached-tokens.md
@@ -0,0 +1 @@
+- **fix(usage):** surface Gemini `cachedContentTokenCount` into `cached_tokens` for non-streaming requests so cache-hit accounting matches the OpenAI/Claude/Responses branches and the streaming path (follow-up to the #10430 envelope fix) ([#10465](https://github.com/diegosouzapw/OmniRoute/pull/10465)) — thanks @rqzbeh
diff --git a/open-sse/handlers/usageExtractor.ts b/open-sse/handlers/usageExtractor.ts
index 313d1ff0e4..f424996ca4 100644
--- a/open-sse/handlers/usageExtractor.ts
+++ b/open-sse/handlers/usageExtractor.ts
@@ -104,6 +104,7 @@ export function extractUsageFromResponse(responseBody, provider) {
     return {
       prompt_tokens: usageMetadata.promptTokenCount || 0,
       completion_tokens: (usageMetadata.candidatesTokenCount || 0) + thoughts,
+      cached_tokens: usageMetadata.cachedContentTokenCount || 0,
       reasoning_tokens: thoughts,
     };
   }
diff --git a/tests/unit/usage-extractor.test.ts b/tests/unit/usage-extractor.test.ts
index 851032690c..af328e52b6 100644
--- a/tests/unit/usage-extractor.test.ts
+++ b/tests/unit/usage-extractor.test.ts
@@ -228,6 +228,7 @@ test("extractUsageFromResponse reads Gemini usageMetadata and thinking tokens",
   assert.deepEqual(usage, {
     prompt_tokens: 11,
     completion_tokens: 7,
+    cached_tokens: 0,
     reasoning_tokens: 2,
   });
 });
@@ -252,6 +253,7 @@ test("extractUsageFromResponse reads Gemini usageMetadata from the antigravity r
   assert.deepEqual(usage, {
     prompt_tokens: 42,
     completion_tokens: 17,
+    cached_tokens: 7,
     reasoning_tokens: 4,
   });
 });
@@ -268,10 +270,34 @@ test("extractUsageFromResponse prefers top-level usageMetadata over the envelope
   assert.deepEqual(usage, {
     prompt_tokens: 1,
     completion_tokens: 2,
+    cached_tokens: 0,
     reasoning_tokens: 0,
   });
 });
 
+test("extractUsageFromResponse surfaces Gemini cachedContentTokenCount as cached_tokens", () => {
+  // Review follow-up on #10430: match the OpenAI/Claude/Responses branches and
+  // the streaming path (usageTracking.ts) by surfacing Gemini cache-hit tokens.
+  const usage = extractUsageFromResponse(
+    {
+      usageMetadata: {
+        promptTokenCount: 30,
+        candidatesTokenCount: 10,
+        thoughtsTokenCount: 3,
+        cachedContentTokenCount: 12,
+      },
+    },
+    "gemini"
+  );
+
+  assert.deepEqual(usage, {
+    prompt_tokens: 30,
+    completion_tokens: 13,
+    cached_tokens: 12,
+    reasoning_tokens: 3,
+  });
+});
+
 test("extractUsageFromResponse returns null when usage is missing", () => {
   const usage = extractUsageFromResponse(
     {